Higher laser power improves strength but reduces corrosion resistance of Mg WE43 processed by powder bed fusion

Powder bed fusion – laser beam (PBF-LB) of Mg alloys provides new possibilities for the production of complex structures with optimized designs, both for weight reduction in aerospace applications, as well as for patient-specific implants in orthopedic applications. However, even though numerous studies have been carried out on the topic, the influence of the individual PBF-LB process parameters on the microstructure and resulting material properties of Mg alloys remains ambiguous. Thus, this study aims to investigate the influence of laser power on the surface roughness, microstructure and resulting key material properties, namely corrosion resistance and mechanical performance. Samples were produced by PBF-LB from gas atomized Mg-4%Y-3%Nd-0.5% Zr (WE43) alloy powder, using three different laser powers: 60 W, 80 W, and 90 W. Contrary to expectation, the 90 W samples exhibited the highest degradation rate, while 60 W samples had the lowest, despite the latter having highest surface roughness and large internal pores. The higher degradation rate for the 90 W samples was instead found to stem from the near-surface microstructure. The higher energy input and subsequently reduced grain size, resulted in an increased amount of second phase precipitates than for the 60 W samples, thereby increasing the tendency for pitting via microgalvanic corrosion. For the tensile strength and elongation at break, the opposite trend was observed. Here, a reduction in grain size and an increase in precipitates for the 90 W samples were found to be beneficial. In conclusion, a definite influence of laser power on the formation of microstructure was observed, ultimately impacting the resulting corrosion and tensile properties of WE43. Future work should investigate the influence of other PBF-LB process parameters, with the aim of establishing an optimum balance between corrosion resistance and mechanical properties.


Introduction
Powder bed fusion -laser beam (PBF-LB) is an advanced metal additive manufacturing (AM) technique, in which orthopedics has already enabled the commercial production of larger patient-specific implants with complex lattice designs for improved bone ingrowth [1][2][3].In other areas, such as aerospace and automotive, PBF-LB lends itself to innovative designs for enhanced functionality, weight savings, and part consolidation [4].
Appertaining to this, Mg and its alloys have been gaining increasing attention in the last 20 years, both for their use in biodegradable orthopedic implants and as lightweight materials in the aerospace and automotive industry [5,6].A Mg alloy was the first metal to be approved for clinical use in a biodegradable metal implant.The implant was an orthopedic fixation device based on a powder extruded Mg-Y-Nd-Zr (WE43) alloy [5,6].However, the commercially available Mg-based orthopedic devices still remain limited to various designs of bone screws.Developing process parameters for PBF-LB of Mg alloys would result in a unique possibility for the development of larger, more complex Mg-based orthopedic implants [7,8].Moreover, due to EU regulations aiming to reduce CO 2 emissions by 55 % as compared to the levels of 1990, the interest in lightweight materials and structures has increased drastically.With the WE43 alloy family also in use as a lightweight material in aerospace, combining the low density of the alloy (1.8 g/cm 3 ) with the design possibilities of PBF-LB, the next generation lightweight structures could be developed.
PBF-LB of Mg alloys was first proven feasible by Ng et al. in 2010 [9].Since then, various studies on the microstructure [10,11], mechanical properties [11][12][13] as well as corrosion properties [14][15][16] of Mg alloys processed by PBF-LB, including the WE43 family of alloys, have been published.WE43 alloys, in particular, show good PBF-LB processability [17,18] with tensile strengths matching those of powder extruded material [11,12].However, although the biocompatibility remains good, and surface treatments demonstrate the capability to hinder the initial onset of corrosion [16], the corrosion rates of the bulk material remain too high for orthopedic applications [7,16,19].The corrosion properties are also the main hindrance in the implementation of Mg alloys in the aerospace and automotive industry, in general [20].In addition, heat treatments have thus far been unsuccessful in notably improving corrosion rates in long-term studies [21][22][23].
Decisive factors influencing the corrosion rates of PBF-LB processed alloys in general have primarily been attributed to microstructure, but also surface roughness and near-surface porosity [24].In Mg alloys in particular, the size and distribution of secondary intermetallic phases, is of major importance.As these phases have a different electrode potential than the Mg matrix, they act as local cathodes, causing microgalvanic corrosion [20,25,26].However, the intermetallic particles can also increase the alloy strength by precipitation strengthening [27].Thus, determining the optimal amount and distribution of precipitates to strike a balance between mechanical properties and corrosion resistance is key for Mg alloys.
In PBF-LB, the formation of microstructure, surface roughness, and porosity are determined by the thermal conditions in and around the melt pool, which in turn are related to the choice of the PBF-LB process parameters [28][29][30][31][32]. Nevertheless, knowledge regarding the relationship between the PBF-LB process and the resulting WE43 microstructure and material properties remains limited.Soderlind et al. found that for single tracks produced with different combinations of PBF-LB process parameters, different types of microstructures were formed for a WE43 alloy [33].However, to the best of the authors' knowledge, studies relating individual PBF-LB process parameters to the bulk microstructure, the surface roughness, and final material properties are currently lacking for the WE43 alloy, and understanding these relationships is key in order be able to develop new PBF-LB products.
Thus, this study aims to investigate the influence of laser power on the macro-and microstructure and the resulting material properties of a PBF-LB processed WE43 alloy, with a focus on the characterization of asbuilt samples.The influence of three different laser powers, centered around a previously established optimized value with regard to porosity and dimensional stability, was first evaluated in terms of surface roughness and porosity.The influence on the microstructure was then assessed by characterization of grain morphology and distribution of secondary elements.Finally, these results were correlated to the degradation rate and tensile properties of the alloy, ultimately providing better insight into the desired microstructural features for improved corrosion resistance and mechanical properties for a PBF-LB processed WE43 alloy.

Sample Production
A gas atomized Mg-3.9 wt%Y-3.0wt%Nd-0.49wt%Zr alloy powder with particle size ranging from 25 µm -67 µm (NMD GmbH, Hemseen, Germany) was processed by PBF-LB on an EOS M100 (EOS GmbH, Krailling, Germany).The process parameters were previously optimized with regards to minimizing porosity and maximizing dimensional stability, resulting in a laser power of 80 W, a powder layer thickness of 0.02 mm, a laser scanning speed of 1100 mm/s, and a hatch distance of 0.05 mm.The samples were produced in an argon atmosphere (O 2 < 0.1 %), with a 67 ˚rotation of the laser scanning direction between each layer, and a laser spot size of 45 µm.
In order to fulfill this study's aim, samples were also produced with a laser power of 60 W and 90 W.These laser powers were chosen as there was a visible difference compared to the 80 W samples, with 60 W samples beginning to show a lack of fusion and 90 W samples showing signs of over-melting.Cylindrical samples with a diameter of Ø = 5 mm and a height of 10 mm were printed for the characterization of microstructure, and surface roughness, as well as for corrosion testing.For the mechanical testing, cylindrical samples based on the ASTM E8M 4c20 tensile test bar [34] were printed (Fig. 1).
As opposed to the corrosion properties, the size of the sample could influence the mechanical properties.Hence, the tensile samples were designed to have a gauge thickness close to that of the strut size typically used in PBF-LB printed lattice structures [16,23].The build direction (BD) concerning sample orientation is indicated in Fig. 1.
All samples were sectioned using dry cutting to avoid possible reactions taking place between the surface and any cutting fluids.To avoid overheating of the samples during cutting, a low cutting pressure was used together with an oscillating sample holder.Before further characterization, the samples were washed with ethanol in an ultrasonic bath for 10 min to remove any loose powder particles at the surface.The bottom of the cylindrical corrosion samples, i.e., the part that was attached to the build plate, was ground with abrasive silicon carbide papers down to P4000 grit to remove any burrs left from cutting, and to ensure that the samples were of similar heights.

Macro-and microstructural characterization
The roughness of the as-built surfaces was measured with an Infinite Focus SL 3d measurement system using a 10x objective (Bruker Alicona, Graz, Austria).Surface roughness measurements were carried out on five samples from each sample group, on an area measuring 8 mm * 2 mm.The curvature of the surface was corrected for in the software.The surface roughness is defined as the arithmetical mean height of the protuberances of the surface (Sa).
For the porosity and microstructural analysis, the samples were mounted in conductive Bakelite and ground with abrasive silicon carbide papers of grades P600, P1200, P2500, and P4000, and subsequently polished using oxide polishing suspension (OPS) (OP-U, Struers, Copenhagen, Denmark) for 15 minutes.The microstructure was first examined using a light optical microscope (LOM) (Leica DM IRM, Leica Microsystems GmbH, Wetzlar, Germany) after 5 s of etching with 2 % Nital (methanol with 2 wt% nitric acid).Characterization of the microstructure was also done by scanning electron microscopy (SEM) (Sigma 300, Zeiss, Oberkochen, Germany) using backscatter electron imaging (BSE) and electron backscatter diffraction (EBSD) (hkl Detector, Oxford Instruments, Abingdon, UK).For EBSD, the electron beam was set to 15 kV and 2.8 mA, and a step size of 1.35 µm was used.The analysis of the EBSD images was done using Aztec Crystal (Aztec 5.0, Oxford Instruments, Abingdon, UK).Grain diameter was defined by the maximum Ferret diameter [35], while grain size was defined as the surface area (µm 2 ).
To quantify the secondary phases that were present in different areas of the samples, image analysis of BSE -SEM images was carried out using Fig. 1.Tensile test specimen dimensions.Dimensions were based on the ASTM E8M 4c20 standard [34].Build direction (BD) was parallel to the length of the sample, as indicated by the arrow in the bottom right corner.

H.N. Åhman et al.
ImageJ [36].This was possible given the contrast differences in BSE mode between the heavier rare-earth elements (i.e., Y and Nd) and the lighter Mg matrix, along with the high concentration of Y and Nd in secondary phases as compared to the Mg matrix.Thus, to determine the amount of secondary phases present, ten images were taken from various regions of the same cross-section using BSE-SEM.The contrast, brightness and working distance were kept constant.The images were subsequently processed in ImageJ [36] to determine the area percentage of secondary phases.In ImageJ, the images were then converted into binary images, setting the threshold value to 60 for all images, and the analysis tool was used to calculate the area percentage, according to the instructions by the software provider.All images are displayed in the transverse plane, with the build direction pointing upwards, unless otherwise stated.

Corrosion
Corrosion testing was carried out on the as-built surfaces using a corrosion media consisting of a Dulbecco's phosphate buffered saline solution (DPBS) (Sigma Aldrich, St. Louis, MO, USA).The composition of salts in the DPBS, given in Table 1, is similar to that found in the human body.The ratio between the corrosion medium and the sample surface area was above 20 ml/cm 2 , in accordance with ASTM G31 [37].The corrosion behavior was characterized for four of the samples per group used for the surface roughness measurements, measuring the H 2 evolution of the samples over 28 days, using a volumetric method as described by Song et al. [38].

Tensile test
The influence of laser power on the mechanical properties of the asbuilt samples was evaluated by tensile testing, until failure.The tests were performed according to ASTM E8M [39], using a Shimadzu AGS-X universal tensile testing machine (Shimadzu, Kyoto, Japan).The load cell capacity was 5 kN and the samples were tested at a crosshead speed at 1 mm/min.Stress was calculated as the force measured by the load cell divided by the cross-sectional area of the specimens derived from caliper measurements.The strain was calculated as the elongation of the specimen over the gauge length divided by the gauge length.Yield strength (σ Y ) and ultimate tensile strength (σ UTS ) were calculated from the engineering stress-strain response for each specimen.To allow estimation of strain at failure (ε), two lines were carefully inscribed on the samples, and the distances were measured before and after the test.The Young's modulus was estimated from the cross-head displacement given by the machine.

Statistical analysis
To determine if there were any statistical differences between groups (60 W, 80 W, and 90 W samples) in terms of surface roughness, tensile yield strength, ultimate tensile stress, and elongation at break, statistical analysis was performed using GraphPad Prism (version 8.0.0 for Windows, GraphPad Software, San Diego, California USA, www.graphpad.com).As all groups displayed a normal distribution, parametric statistical analysis in the form of a one-way ANOVA was carried out, using the Tukey post-hoc test to establish any significant differences between groups.The difference in the number of secondary phases present between the two types of microstructural regions observed, i.e. cellular and dendritic, was assessed through an unpaired t-test using GraphPad Prism.

Results and discussion
A typical macroscopic appearance of the as-printed cylindrical samples for the different powers used is shown in Fig. 2a) -c).The surface of the 60 W samples appeared rougher and slightly darker than those of the 80 W and 90 W samples, while no difference between the 80 W and 90 W samples was evident from visual inspection.Moreover, the edges of the 90 W samples appeared slightly more rounded and shinier than for the other two.Evaporation products were observed during printing for all the samples, but no visible difference in the amount of evaporation was seen.Evaporation of Mg is often reported [15,40] and is also expected due to the small difference between melting temperature (650 • C) and boiling temperature (1090 • C) [41].
Regarding the surface roughness, a statistically significant difference was found between the 60 W and 90 W groups (p<0.05) (Fig. 2d)) Comparing the absolute results of surface roughness with those found in other publications is difficult, as the method of measuring varies, and this influences the results [42].Nevertheless, for other alloys processed by PBF-LB, the as-built surface roughness as measured with an Alicona Infinite Focus ranges between 5 µm -40 µm; in agreement with that for our samples [43][44][45].
For other alloys, it has been established that the surface roughness is influenced by powder size, layer thickness, and hatch distance, which in this case was the same for all samples.It has also been shown that an increase in laser power or a decrease in laser scanning speed can, within a certain range, decrease the surface roughness, while outside this said range, the opposite result is achieved [46].The reason for obtaining a higher surface roughness for the 60 W samples than for the 90 W samples could be related to the lack of energy input for the 60 W samples, resulting in a lack of fusion, and a higher number of partly melted powder particles stuck to the surface.Indeed, as can be seen in the LOM images, the 60 W samples (Fig. 3a and d) feature large pores in the size range of 100 µm -500 µm.The size and shape of the pores indicate that they are formed due to lack of fusion, i.e., there was not enough energy input to melt the material [47,48].In the 80 W and 90 W samples, there were no lack-of-fusion pores present, but a few smaller pores in the size range of 10 µm -20 µm, typically related to keyhole formation or porosities present in the powder, originating from the powder production [47,48].In this case, it could be either, as pores have previously been observed in the powder used in this study [17].After etching, dendritic regions can be observed along the edges of all the samples, indicated by the squares in Fig. 3d)-f), but that this region goes deeper into the 90 W samples in comparison to the 60 W samples.The dendritic structure can be seen at higher magnification in Fig. 4, often seen in various types of Mg alloys [41].
There is also a clear difference in the morphology of the top layer for the three samples.In the 60 W samples (Fig. 3a)) the top layer is more undulating with the right corner protruding, also indicating an unstable surface related to the lack of fusion.In the case of a large build, this  could lead to damage to the recoater and defective builds.For the 90 W samples, the rounding of the top right corner could be related to the rounding of the edges seen during the visual inspection of the samples.The rounding of the edges is usually related to the accumulation of heat along these regions due to an isolating effect of the powder [49].These edges have the potential to develop into protruding ridges for bigger builds and could just as for the 60 W samples lead to damage to the recoater and defective builds.Finally, from the last layer, it can also be seen that the melt pools are shallower for 60 W samples than for 80 W and 90 W samples, as expected due to the lower energy input.The depth of the melt pools in the 90 W samples is within the same range as the 80 W samples, while the width cannot be distinguished in these images.Moreover, dendritic structures are more prominent in the top layer for the 80 W and the 90 W samples (Fig. 3a) and b)), while none are visible for the 60 W samples (Fig. 3c)).
Further examination of the material by BSE-SE, identified two types   of microstructural regions in the samples, namely, cellular (Fig. 5a)) and dendritic (Fig. 5b)).In Fig. 5a) a cellular structure is mainly present within the melt pool, separated by bands of RE-rich precipitates arranged parallel to the melt pool (2).Along the melt pool boundaries (1), there is also a cellular structure present [50], orientated perpendicular to the melt pool boundary (3) [10].Fig. 5b) shows regions of mainly a dendritic structure, with the alpha Mg dendrites (4) clearly distinguishable by their six-arm formation [33,51].Both dendrites, cellular and lamellar structures are separated by Mg-RE segregates.The cellular structures observed in Fig. 5a), are also present in areas between the dendrites in Fig. 5b).Similarly, formation of dendrites is also seen in Fig. 5a).Oxide flakes (5) are present throughout the microstructure in both cellular and dendritic regions (5).Although these structures are present in all three samples, the dendritic structure was more prevalent close to the edges of the 90 W samples (Fig. 6), while the cellular structures were more prevalent in the 60 W samples, which is in line with observations made in the LOM images in Fig. 3.Moreover, studying the BSE-SEM images along the top edge of the samples, i.e., the last layer melted, mainly a cellular structure can be seen for the 60 W sample (Fig. 6a) while for the 90 W sample mainly a dendritic structure is present.Dendrites are also present to a higher degree in the bulk of the 90 W sample.Another difference between the samples is the morphology of the oxides, being present as larger flakes in the 60 W sample.
The oxide flakes and the Mg-RE intermetallic compounds are discussed in most other studies on PBF-LB of WE43 alloys [18].The oxide flakes are generally said to originate from the thin oxide layer present on the surface of the powder, while the Mg-RE intermetallic compounds either precipitate on a preferred plane [10,52], or segregate at the dendrite and cellular boundaries [51,53].However, the occurrence of the microstructural features such as the cellular and dendritic structure, as well as the denomination of the same, varies between studies [16,17,54,55].Moreover, the difference in the morphology of the oxides has not been previously discussed.
Soderlind et al. [33] investigated the effect of laser power, laser scanning speed as well as spot size on the aforementioned microstructural features, but only for single tracks.They concluded that depending on the thermal conditions in the melt, mainly dictated by the spot size, either cellular, dendritic, or planar microstructures were formed.Moreover, Soderlind et al. also complemented the experimental analysis with computational thermodynamic analysis based on the CALPHAD methodology, and the cellular and planar modes were predicted, while the dendritic structure was not.The dendritic structure only occurred when a smaller spot size of 50 µm was used, as opposed to 80 µm.Moreover, it mainly occurred for laser powers from 200 W and above, with the exception of also occurring for 150 W when combined with the lowest scanning speed, corresponding to 75 mm/s.They argued that a higher heat input was the main reason for the dendritic formation, given the increased heat input attributed to smaller spot size and higher laser power.However, they also hypothesized that the non-equilibrium keyhole-mode condition was a contributing factor.Moreover, the reason that the dendritic structure was not predicted by the thermodynamic model was ascribed to its shortcomings with regard to application for non-equilibrium conditions.
Bär et al. [10] also observed the formation of dendrites in the center of the melt pools, while a cellular structure was seen along the melt pool boundaries.This was attributed to the lower temperature gradient in the middle of the melt pool, whereby dendrites typically form for low-temperature gradient regions [56].Relating these studies to the results of this work, it can be hypothesized that the reason for a higher number of dendrites occurring along the edges of the samples is because of the insulating factor of the powder due to its low heat conductivity [57,58].As there is less amount of material dissipating the heat, this leads to a warmer localized area and slower cooling rates (i.e., low-temperature gradient).In turn, this could also result in unstable melt pools.Similarly, the reason for the dendritic structure being more prevalent for the 90 W samples, both in the bulk and the last layer melted, could thus either be due to the higher energy input, which could either enhance non-equilibrium melting conditions or result in an even lower temperature gradient due to the buildup of heat over several layers in the printed part, allowing the dendrites to form.A warmer process could also result in a higher degree of evaporation of Mg at the edges, thereby increasing the alloying content locally, and thus increasing the precipitation and segregation of phases.
A difference in the morphology of the oxides in the samples produced with 60 W and 90 W was also observed.This could be explained by the higher laser power applied for the samples produced with 90 W, resulting in a higher movement in the melt pool breaking down the oxide flakes Fig. 7.It could also be due to the higher energy of the 90 W laser in itself, breaking the oxides present in the powder particles into smaller fragments on impact, or due to a higher evaporation pressure of the melted Mg.Smaller oxides are sometimes added to Mg alloys for grain refinement, resulting in dendritic grains [59,60], so the smaller oxides may contribute to the formation of the dendritic structure.However, the distribution of oxides and its impact on the formation of the different types of microstructures observed needs further investigation.
In addition, the image analysis of the BSE -SEM images shows a significantly higher area fraction of secondary phases in the dendritic region as compared to the cellular region.Representative BSE-SEM images used for quantifying the amount of secondary phases in the cellular and dendritic regions are presented in Fig. 8a) and b), with the images after processing by Image J presented in Fig. 8c) and d).A statistically significant difference was found in the number of secondary phases present, corresponding to an average of 17.6±1.7% in the cellular regions, and 24.6±1.6% in the dendritic regions.
The crystal orientation maps obtained from EBSD measurements of the near surface region along the walls of the samples are presented in Fig. 6.SEM images of the near surface microstructures of the samples built with a) 60 W b) 80 W and c) 90 W.

H.N. Åhman et al.
Fig. 9, with the grain size distribution plotted against area weighted fraction plotted in Fig. 9. Two types of grain structures can be observed (Fig. 8).Firstly, the smaller equiaxed grains measuring a few µm in diameter, i.e. areas marked 1 in Fig. 8, had a weaker texture (Fig. 11).Secondly, there were larger grains measuring up to 150 µm in diameter, i.e. areas marked 2 in Fig. 8, that had a strong basal texture (Fig. 11).To further highlight the strong basal structure of the larger grains, the area marked 3 in Fig. 8 has the IPF coloring with the 0001-direction parallel to the transverse direction.These grain structures have also been previously observed for PBF-LB processed WE43 [14].For the sample produced with 60 W, the equiaxed grains are only situated along the very edge, reaching a little more than 100 µm into the material.For the 90 W, however, the small equiaxed grains are present more than 500 µm into the material.The increased amount of small equiaxed grains along  the edges of the 90 W sample is in line with the dendritic structure observed by BSE-SEM in Fig. 5 and Fig. 6, and LOM in Fig. 3.
These results are in line with previously mentioned studies [10,33].Soderlind et al. also observed the small equiaxed grains for samples having dendritic microstructure.Likewise, Bär et al. [10] observed the equiaxed grains in the center of the last melt pool, with elongated grains along and perpendicular to the melt pool boundaries.However, Jauer et al. [40], as well as Zumdick et al. [11] both, mainly observed the small equiaxed grains in the bulk of the sample, while Esmaily et al. [22] mainly saw the large basal grains in the bulk of the material for all of their three sets of optimized process parameters [15].The reason for the difference in reported microstructure is not yet established, but it has previously been suggested that the difference in grain size distribution could be related to the alloying content [18].However, both Zumdick et al. [22] and Esmaily et al. [22] printed a WE43 alloy on an Aconity MINI, using a layer thickness of 30 µm, a hatch distance of 40 µm and a laser spot size of 90 µm, but a different combination of laser power and scanning speeds.Nevertheless, as previously discussed, Soderlind et al. [33] showed that different microstructures can occur (i.e., planar and cellular grains), depending on the laser setting.Moreover, this study shows that both grain structures can be present, as a function of the local thermal conditions, related both to the location in samples as well as the laser power applied.
The degradation study results showed a large variation in H 2 evolution within each group (Fig. 12), indicating an inhomogeneous material.Furthermore, the 60 W sample had the slowest degradation rate, despite the higher surface roughness and large pores.However, the pH increase during the corrosion tests did not differ significantly between the samples (Fig. 12 b).Any increase in pH is important to note, as this typically slows down the degradation rates [20].Moreover, an increase in pH is also detrimental in vivo, although in a living system, there are mechanisms counteracting the increase in pH [61].Future studies should include dynamic corrosion studies to better mimic in vivo conditions.
After 28 days of immersion (Fig. 13), the corrosive attack was most severe on the top and bottom surfaces of the cylindrical samples, with the bottom part being the part attached to the build plate, and the top part being the last printed layer of the sample.The attack on the bottom of the sample could be explained by any internal defects such as pores being exposed after the samples were cut from the build plate.In contrast, the attack on the top, more prevalent for the 90 W sample, could be related to a large amount of secondary phases attributed to the dendritic structure also present in the last melted layer (Fig. 7).Thus, the large pores in the 60 W sample did not have a decisive effect on the degradation, possibly due to i) them being mainly internal pores, and ii) the 28-day degradation being too short to reach the areas where these pores are present.The lack of effect of the surface roughness is thus also due to other factors playing a more detrimental role to the degradation rate, such as the distribution of the secondary phases.In general, the main factor determining the corrosion properties of Mg alloys has been found to be the amount and distribution of secondary phases [20,27,62,63].This is due to the large difference in electrochemical activity between the secondary phases and the Mg matrix, causing the secondary phases to act as local cathodes, resulting in microgalvanic corrosion [64].
These results suggest that the dendritic structure is more detrimental to the corrosion properties than the coarse basal grains.This could be Fig. 9. EBSD orientation maps with the IPF coloring set with the 0001-direction parallel to the build direction of the edges of samples produced with a) 60 W, b) 80 W and c) 90 W. Areas encircled corresponds to 1) dendritic grains, and 2) large basal grains.To highlight the strong basal texture of the large grains, insert 3 in a) shows the basal region with the IPF coloring set with the 0001-direction perpendicular to the build direction.due to the size and distribution of secondary phases along the dendritic regions.It could also be due to a higher degree of evaporation of magnesium, and thus a shift in the chemical composition of alloying elements for the 90 W. For instance, a decrease in Mg content due to evaporation could increase the electrochemical potential between the secondary phases and the alpha Mg matrix, leading to an increased likelihood of microgalvanic corrosion.Moreover, there are some studies showing a difference in corrosion rate related to texture, but as the 90 W sample mainly has dendritic grains in the last layer, which typically have a random texture, this is unlikely to have had a major influence on these results [65,66].In order to eliminate the impact of the corrosion attack on the top and bottom surfaces in future tests, these regions could be covered by silicon during testing.Nevertheless, the difference in corrosion attack in different parts of the printed parts is also a problem that has to be solved in the final application, for example by surface treatments.
The lack-of-fusion pores seen in the corrosion samples for the 60 W group were not present for the tensile test samples printed with the same set of process parameters (Fig. 14).However, for the 80 W and 90 W samples, larger, spherical pores were now visible.The size of the pores, some of them being larger than the particle size of the powder, indicated that they are keyhole porosity.The change in porosity in all three types of samples can be explained by the thinner cross-section of the tensile test samples in comparison to the corrosion samples.Indeed, for thinner samples, there is a shorter laser scan length and the insulating effect of the powder surrounding the samples is more prominent.This leads to the disappearance of the lack-of-fusion pores in the 60 W samples, while in the 80 W and 90 W samples, keyhole pores appear.However, as was discussed for the corrosion samples, there is a greater prevalence of regions with smaller grains throughout the 90 W samples as compared to the 60 W samples, again with dendritic grains being present further into the material.The keyhole porosity also mainly appears in these regions.Although the difference in sample size results in some changes in the internal porosity of the samples, the larger amount of dendrites along the edges are still present for the 90 W sample. EBSD orientation maps (Fig. 15) corresponding to the areas encircled in Fig. 14 confirm that the regions indicated are dendritic.[68,69] Regarding the mechanical properties including yield strength (σ Y ), ultimate tensile strength (σ UTS ) and strain at failure (ε f ), a statistically significant difference was found for all three measurements between 60 W and 90 W (p<0.05)) (Fig. 16 and Fig. 17).The 60 W sample had the lowest σ Y , σ UTS , as well as the lowest ε f , while the 90 W samples exhibited the highest σ Y , σ UTS, and ε.There was also a statistically significant difference between 80 W and 90 W (p<0.05) for all values, but no statistically significant difference between the 60 W and the 80 W samples for neither σ UTS nor ε f .A Young's modulus of around 40 GPa was obtained for all samples, which is in agreement with the literature [41].
The values obtained for σ Y and σ UTS exceed the values reported for cast WE43 (σ Y-cast ~160 MPa, σ UTS-cast ~200 MPa) but are slightly lower than what has previously been reported for PBF-LB samples [11,12,67].However, the samples in this study are smaller, and have as-built surfaces, resulting in a higher surface roughness.The importance of surface defects on the tensile properties of PBF-LB processed alloys, including factors such as surface roughness and near-surface defects, have previously been established [46,68,69].Moreover, the low thickness of the samples enhances the importance of the surface, due to the lower bulk-to-surface ratio [22].The surface roughness could also lead to an overestimation of the cross-section, and thus an underestimation of the mechanical property values [70].Finally, although the ε f remains equal to or higher than what is generally reported for cast material (ε f-cast =4 %   -6 %), and is in line with what has previously been recorded for PBF-LB-processed WE43, the ε f is much lower than for wrought WE43 for all three specimens [11,12,67].Nonetheless, these results show the feasibility of printing thin structures while maintaining decent tensile properties.
To further investigate the reason for the difference in tensile behavior between the samples, the fracture surfaces were characterized.For both the 80 W and the 90 W samples the pores observed in the LOM images in Fig. 14 were clearly visible also on the fracture surfaces in Fig. 18.The fracture surface of the 60 W sample showed more signs of localized cleavage planes than the 90 W sample, indicative of a brittle fracture.This is in accordance with the elongation at break values (Fig. 17).
The lower tensile strength and the more brittle behavior of the 60 W samples was unexpected, as porosity generally has an exponentially negative influence on the mechanical properties and more keyhole porosity was observed in the other groups [12].Thus, other mechanisms were clearly more important here.Many Mg alloys typically exhibit a low yield strength and a brittle type of fracture due to the limited amount of activated slip systems at room temperature.The increase in yield strength of rare earth (RE) containing alloys is generally related to the solid solution of the alloying elements and the formation of Mg-RE intermetallic compounds, blocking dislocation movements and suppressing intergranular deformation [71][72][73][74], while the increase in ductility is generally due to grain refinement and texture weakening [75,76].Both mechanisms would explain the increased strength due a larger amount of secondary phases in the 90 W samples.Moreover, as the higher amount of dendritic grains results in a finer grain structure in the 90 W, the Hall-Petch effect could also be a contributing factor to higher strength [77,78].Another factor could be the size and distribution of the oxides.As mentioned previously, oxide particles have been added for grain refinement in Mg alloys, but they can also be added to act as a strengthening phase in oxide dispersion-strengthened alloys, filling the same function as Mg-RE intermetallic compounds, i.e. blocking dislocations and suppressing intergranular deformation [59].However, these mechanisms are only at play if the strengthening phase is of the right size, and large particles could lead to an opposite effect.Thus, it could also be that the larger oxides in the 60 W sample had a weakening effect [60].The reason for finding no significant difference between the 60 W and 80 W samples in σ UTS and ε UTS could be that there is a threshold for the amount of smaller grains, and thus texture weakening, that is passed when applying 90 W.
Comparing the corrosion and the tensile test results, it is evident that while the 60 W sample exhibited the highest corrosion resistance, it also presented the lowest tensile strength, as expected.As previously discussed, the effects generally giving rise to the increase in mechanical properties, such as precipitation of secondary phases, resulted in a decrease in corrosion resistance.While the top and bottom surfaces resulted in an important contribution to the corrosion results, but had no part in the tensile test results, this highlights the importance of the local   microstructure with regard to the material properties for PBF-LB processed WE43 alloys.Moreover, these are the first results that give insight into the importance of laser power alone in the formation and role of the different microstructures observed for PBF-LB processed WE43 alloys, as well as its effect on the material properties.It is also the first study to highlight the importance of wall thickness for this alloy, which is of utmost importance considering the intended application of lattice structures, which might include both thinner structures and varying wall thicknesses.Future work should investigate the effect of other process parameters beyond that of laser power on the microstructure of PBF-LB processed WE43.As PBF-LB is a highly complex process, where not only the local thermal conditions can have an influence, but also melt pool dynamics play an important part in the microstructures formed [33].Further clarification of the mechanisms behind the formation of the various microstructures observed, both locally and globally, as well as their influence on material properties, is needed to be able to develop complex components with sufficient corrosion resistance and mechanical properties for applications within both aerospace and medical technology.Moreover, enhancing the understanding of the relationship between the process itself and final part properties is of utmost importance for part qualification and quality control in both aerospace and medical technology.
Important factors for end-use applications include the geometry and thickness of the parts, particularly with regard to the design of patientspecific implants and complex lattice structures.Such aspects are crucial for the successful production of complex lattice structures for use as e.g.biodegradable implants, with controlled degradation rates and suitable mechanical properties.

Conclusions
In this study, the macro-and microstructure of a WE43 alloy processed by PBF-LB using three different laser powers, 60 W, 80 W, and 90 W, were investigated and related to the resulting tensile and corrosion properties.From the results, it can be concluded that a change in laser power had a significant effect on both aspects.
Specifically, an increase in laser power from 60 W to 90 W resulted in a lower surface roughness, but an increase in dendritic grain structure, and thus the amount of secondary phases present.The change in surface roughness was related to a lack of fusion at the edges, for the 60 W samples, while the increase in the dendritic grains in the 90 W samples could be related to an unstable melt pool.In the corrosion samples, lack of fusion pores was observed for the samples produced with 60 W, while for the tensile test specimens, keyhole porosities were observed for the samples produced with 80 W and 90 W.Although the change in sample size affected the pore formation, the near-surface  microstructure was similar for both sample groups.
The increase in laser power from 60 W to 90 W resulted in a decrease in corrosion resistance, while the mechanical properties in terms of tensile yield strength, ultimate tensile strength, and elongation at failure, increased.The increase in yield strength and ultimate tensile strength was explained by the Hall-Petch effect and precipitation strengthening due to the increase in dendritic structures, with smaller grains and a higher amount of secondary phases.The smaller size of the oxides may have also contributed to the improved tensile properties for the 90 W samples.
Although the change in sample thickness resulted in a change in the size and distribution of the pores, the main contribution to both corrosion and tensile properties were found to be related to the dendritic microstructure and the distribution of secondary phases.
Finally, this study shows the feasibility of printing thinner structures, while maintaining tensile properties that are better than cast material.It also highlights the need for further studies looking at the optimization of the PBF-LB process with regard to microstructure formation and its relation to material properties, in particular corrosion resistance.

Fig. 2 .
Fig. 2. Photos of representative samples of the three sample groups being produced with a laser power of a) 60 W, b) 80 W and c) 90 W, with the white square in a) showing an example of the location of the surface roughness measurements.d) Results for the mean surface roughness, Sa (µm), as measured with an optical measurement technique (Alicona Infinite Focus, n=5 / group).

Fig. 4 .Fig. 5 .
Fig. 4. LOM image showing a dendritic region in a sample produced with 90 W at higher magnification.

Fig. 7 .
Fig. 7. BSE-SEM images of the microstructures observed in the top edge of the samples, i.e. the last melted layer of the samples in comparison to the bulk of the samples, images a) and c) corresponding to samples produced with 60 W, and images b) and d)corresponding to samples produced with 90 W.

Fig. 8 .
Fig. 8. BSE-SEM images used for quantifying the amount of secondary phases in cellular (a and c) and dendritic (b and d) regions, including images before (a and b) and after (c and d) processing by Image J. The images show a statistically significant higher area fraction of white (corresponding to the secondary phases) for the dendritic areas as compared with the cellular regions.

Fig. 10 .
Fig. 10.Grain size as defined by surface area plotted against area weighted fraction.

Fig. 11 .
Fig. 11.Pole figures corresponding to the subsets of grains having a surface area that is a) larger and b) smaller than 100 µm 2 in the IPF color map Fig.8c).The build direction is orientated out of the image, parallel to <0001>.

Fig. 12 .
Fig. 12. Degradation test quantifications of a) hydrogen evolution and b) change in pH.

Fig. 13 .
Fig. 13.Macroscopic appearance of the samples after 28 days of immersion.

Fig. 14 .
Fig. 14.LOM images of the cross-section of the waist of the tensile test samples for a) 60 W and b) 80 W and c) 90 W. The arrows indicate regions with dendritic grains, and the encircled areas correspond to the regions where EBSD data was collected.For IPF color maps are presented in Fig. 12.

Fig. 15 .
Fig. 15.EBSD orientations maps of tensile test samples produced with a) 60 W b) 80 W and c) 90 W. The two lines appearing in the 80 W samples are scratches originating from the sample preparation.The IPF coloring is set with the 0001-direction parallel to the build direction.

Fig. 16 .
Fig. 16.Representative stress-strain curves for the samples produced with the different laser power (n=5 per group).

Fig. 17 .
Fig. 17. Results from the tensile tests, showing a) yield strength, b) ultimate tensile strength c) elongation at ultimate tensile strength.

Fig. 18 .
Fig. 18.SEM images in various magnifications of the fracture surfaces for samples produced with 60 W (a), d) and g)), 80 W (b), e) and h)) and 90 W (c), f) and i)).

Table 1
Salt concentration of the Dulbecco's phosphate buffered saline solution (DPBS) used in the corrosion experiments.